Greek Personal Names and Linguistic Continuity

ANNA MORPURGO DAVIES

The linguistic study of personal names

THE STUDY OF PERSONAL NAMES does not always arouse enthusiasm among historical linguists. Most onomastic work is seen as etymological in nature and consequently flawed, since personal names are even more difficult to ety- mologize than normal nouns. Yet it is easy to identify important linguistic work on the subject that is not etymological and carries a high degree of linguistic credibility. A crucial question concerns the way in which the diachronic development of personal names differs from that of other lexical items. Though at first sight the question seems to be of interest only to the linguist, I want to argue that an answer can be helpful for the historian too. There are risks in the enterprise: on the one hand we require a set of gener- alities which may seem either woolly or dull or both (and will not be new for the linguist), on the other all general statements must be supported with very detailed information which comes close to pedantry.

Personal names vs. lexical items

Historical linguists are trained to study the phonological, morphological, and syntactical changes incurred by words and combinations of words through time. They also study meaning changes, i.e. they look at the contexts in which words occur in different periods and from those contexts extract

Proceedings of the British Academy, 104, 15–39. © The British Academy 2000.

information which allows them to establish the meaning of the word and, if possible, to date the various stages of its semantic development. A linguist will be interested, for instance, in observing that the Old English form of the word ‘soon’ was so¯na and that it meant ‘at once’, rather than ‘soon’. Here we meet a first contrast with the study of personal names. It would be odd to ask in so many words whether the meaning of the name Elizabeth has changed between the time of Elizabeth the First and that of the present queen. On the other hand it is reasonable to ask how the name Elizabeth was pronounced and/or spelled in the sixteenth century. In other words, though we cannot discuss in the same way the meaning and meaning change of personal names and common nouns, the phonological shape of names norm-ally changes together with the phonological shape of other elements of the lexicon. Here too, however, there are differences. We normally expect any sound or sound sequence to alter in a way which is paralleled in other words: if Old English so¯na [so:na] becomes soon [su:n], we also expect to find that OE mo¯na [mo:na] and OE fo¯da [fo:da], if continued, become moon [mu:n] and food [fu:d], as they in fact do. The pattern is clear and regular and it involves replacement. Different individuals may alter their pronunciation in different ways and at different speed, but normally, in the same language or dialect, after a period the old and the new form do not co-exist: we do not say both [so:na] and [su:n]. But if we look at an Old English name like Æðelpı ryd, by the end of the nineteenth century this seems to be continued by three different names: Audrey, Ethel, and Etheldreda. We do not find this altogether surprising, but we would think it much odder if OE so¯na was continued by e.g. *so, soon and *sonar.

Structure and evolution of personal names

Personal names may differ from other lexemes both in their structure and in their evolution pattern. Before asking why this is so we ought to explore some of these differences. The best way to do it is to look at real examples (some of which may look unbearably trivial) and consider what they teach us. My exem- plification is based on Greek and starts with questions of word-formation. Consider first two adjectives, both found in the same late Thessalian inscription (IG IX (2) 517, Larisa, late third century BC): λθιοy ‘made of stone’1

1 In Attic, where the equivalent of Thessalian λθιοy is λθινοy, adjectives of material are formed either with an inherited *-eyo- suffix (cf. χρυσο y < χρυσ οy) or with an -ινο- suffix as in λθινοy.

and Ενµειοy ‘son of Eunomos’; they are both synchronically derived from thematic words (λθοy, Ενοµοy) with two productive suffixes, -ιοy and -ειοy. We have every reason to believe that, for the patronymic adjective too, the original suffix was -ιοy, but patronymic adjectives which end in -ειοy and not in -ιοy are not limited to this inscription. In Thessalian in general, -ιοy was replaced with -ειοy in the derivatives from most pers-onal names but not in the derivatives from common nouns. Secondly, consider frequent names such as Θεδοτοy or ∆ιδοτοy. Both names are wholly transparent and fit in the general category of -δοτοy compounds; indeed we have evidence for a straightforward adjective θεδοτοy ‘god-given’. However, while the -δοτοy compounds are two-termination words, i.e. they do not have different forms for masculine and feminine, next to the personal name Θεδοτοy we find a feminine name Θεοδτη and next to ∆ιδοτοy a feminine ∆ιοδτη. In other words, the standard rules of Greek word-formation are broken for this type of personal name. Similarly, and again in contrast with the standard derivational pattern, we find feminines like Θεοδρα next to Θεδωροy, Γλαυκππη, Ξανθππη next to Γλα!κιπποy, Ξα´ νθιπποy, Καλλκλεια next to Καλλικλ$y, Μενεκρα´ τεια next to Μενεκρα´ τηy, etc. These feminines are old, since in the second millennium BC we find at Mycenae (V 659) te-o-do-ra (Θεοδρα) and a-re-ka-sa-da-ra ( &Αλεξα´ νδρα), and Homer has names like &Αµφιγ νεια and Ερυκλ εια. However, here too compound adjectives of the πολ!δωροy, µελα´ νιπποy, φλανδροy, εγεν+y type do not normally form separate feminines.2 The next question concerns both phonology and morphology. Short forms like Jim for James or Dick for Richard are found in a number of languages. In Greek at all stages, from Mycenaean to the koine, compound names can be replaced by abbreviated forms: in the Iliad one and the same man, Achilles’ charioteer, is called ,Αλκιµοy (Il., 19. 392; 24. 474, 574) and &Αλκιµ δων (Il., 16. 197; 17. 467 etc.). This is not a standard type of phonological change and Greek does not abbreviate all words in this manner. We have here a shortening process which is typical of personal names: the abbreviated form and the full form may co-exist and be treated as identical and perhaps interchangeable (as Tom and Thomas in English), or may acquire separate status and eventually be treated as different names.

2 In the ancient locus classicus about feminines, Aristophanes (Clouds, 658 ff.) does not refer to this specific point, but comments on the feminine gender of κα´ ρδοποy ‘kneading trough’ and the possible female referent of -λεκτρυν ‘cock and hen’, while mentioning the feminine names Σωστρα´ τη and Κλεων!µη.

There is a link between these Kurzformen and a phonological process which is far more common in names than elsewhere: expressive gemination tends to affect the abbreviated forms rather than the full compound. Totake the standard example always quoted: one and the same person, a fourth-century tyrant of Methymna, is referred to by Theopompus (ap. Athen., 10. 442) as Κλεοµ νηy, but appears in an inscription in which he is honoured by the Athenians (IG II2 284) as Κλ οµιy and is quoted by Isocrates (Ep., 7. 8 f.) as Κλ οµµιy;3 we would not expect to find *Κλεοµµ νηy. The examples provided by Vottero for the geminated forms of Boeotia4 confirm this view: cf. for example, ∆α´ µασσιy (IG VII 1908a), presumably from a ∆αµασι- compound (e.g. ∆αµα´ σιπποy). It is impossible to know if, in actual speech acts, other lexical items underwent gemination with the same frequency, but, if they did, this was not indicated in writing and presumably did not impinge on the standard lexicon. In other words, in phonological processes such as shortening, gemination, etc. personal names behave differently from common nouns. Now comes a question of meaning. Composition is one of the standard means by which Indo-European (IE) languages in general, and Greek in particular, enrich their vocabulary. It is a basic rule of thumb that new compounds yield new meanings and that these are normally predictable from the meaning of the component elements. Greek, as Indo-European, frequently forms names from compounds, but there is a fundamental difference between lexical compounds and onomastic compounds. The latter may over- lap with the former (as in the case of Θεδοτοy) but need not; there are onomastic compounds which would not be acceptable as lexical compounds for the simple reason that they make no sense. Solmsen gives as examples words like Λυσκριτοy or Λυσµνηστοy.5 Masson follows Bechtel in assum- ing that some at least of what he calls ‘noms irrationels’ (Κλεοφο0νιξ is an example) arise because the basic aim is to continue traditional family elements in the name.6 As Solmsen had noticed, Aristophanes’ account in the Clouds (60 ff.) of how Strepsiades’ first son was named fits with this pattern. Strepsiades’ wife wanted a -ιπποy name: Χα´ ριπποy or Ξα´ νθιπποy or Καλλιππδη2, but Strepsiades himself preferred Φειδωνδηy, i.e. a normal

3 See most recently O. Masson, BSL 81 (1986), 217–29 at 223 OGS, 549–61 at 555. 4 G. Vottero, ‘Procédés d’expressivité dans l’onomastique personnelle de Béotie’, in La Béotie antique (Paris, 1985), 403–17. 5 F. Solmsen, Indogermanische Eigennamen als Spiegel der Kulturgeschichte, hsg. und bearbeitet von E. Fraenkel (Heidelberg, 1922), 117. 6 O. Masson, Philologus 110 (1966), 253 f.; Rev. Num. 24 (1982), 22 f. OGS, 88 f., 404 f.

name built on that of his own father Φεδων. The compromise is Φειδιππδηy, not a particularly meaningful compound in spite of Homeric Φεδιπποy (Il., 2. 678) and of the Homeric phrase 4ππων φειδµενοy (Il., 5. 202).7 Another feature of onomastic compounds, as contrasted with lexical compounds, is the reversibility of the order of the elements. Next to Hippostratos (attested in Athens from the sixth century), we have in fourth- century Athens Stratippos.8 Masson pointed out the equivalence of the banal &Ονασ-µβροτοy and the rare fifth-century name Μορτ-νασοy from Thera.9 Here too it remains obscure what the meaning of a compound with reversed order would be, but in fact the question of meaning hardly arises.

Function and use of personal names

The facts listed above are well known; they are rehearsed here in order to provide concrete evidence for the assumption that the history of personal names requires a type of analysis which partly overlaps and partly contrasts with that of other words. But why is this so ? Why do personal names show a different pattern of development from other lexical items? The answer in my view is to be found on the one hand in the role and function of personal names in contrast with the rest of the lexicon, and on the other in the way in which personal names are assigned to individuals.

Identification Names identify individuals, often in situations in which the linguistic context does not help. Even when hearing is difficult, it is unlikely that a sentence like ‘the table has a broken leg’ would be understood as ‘the cable has a broken leg’, but ‘Tim just broke a leg’ could easily be heard as ‘Kim just broke a leg’. If we return to my original question about the different derivation in Thessalian of λθιοy ‘made of stone’ and Ενµειοy ‘son of Eunomos’, we

7 Cf. for the Homeric evidence H. von Kamptz, Homerische Personennamen (Göttingen, 1982), 9 f., 63. 8 For the references see LGPN II, 238, 406. 9 O.Masson, ‘Quelques noms grecs à l’Agora d’Athènes’, in G. Cardona and N. H. Zide (eds), Festschrift for Henry Hoenigswald (Tübingen, 1987), 255–9 at 256 OGS, 567–71 at 568.

may explain how the difference arose through a mixture of purely linguistic facts and functional reasons.10 At an earlier stage, patronymic adjectives and other nominal derivatives must have ended in [-aios] when they derived from a-stems (Μνασαιοy from Μνασαy), [-Cios] when they derived from conso- nantal stems (C any consonant) or thematic stems (* &Ιασνιοy from *&Ια´ σων,*Ενµιοy from Ενοµοy), and in [-eios] when they derived from s-stem (-γ νειοy < *-genes-ios from names ending like Attic ∆ιογ νηy). The early change of Thessalian [ei] to [e¯] led in effect to the co-existence of two suffixes [-e¯os] and [-ios], whose original identity was no longer clear. In patronymic adjectives [-ios] was preserved in [-aios] but replaced by [-e¯os] elsewhere. Why [-e¯os] in preference to [-ios] and why only in patronymics? Two reasons can be mentioned. First, [-e¯os] allowed the creation of unambiguous patronymics to names which ended in -ιοy; hence Thessalian 9Ραδιε0οy from 9Ρα´ διοy, ∆ικαιε0οy from ∆καιοy, etc.; it would have been impossible to add -ιοy to something which ended in -ιοy. Secondly, and cru- cially, in Thessalian a sequence such as [-Cios] through a palatalization process tended to be replaced by [-CCos] or [-Cos] while [-Ce¯os] was preserved. In other words, phonological attrition made the original patronymic formation opaque so that the basic name was not always recoverable. In normal words, the need for unambiguous formations was far less strong since the context could disambiguate the form—hence the preservation of -ιοy in λθιοy.

Classificatory function The primary role of personal names is to identify individuals, but personal names are also endowed with a strong classificatory function: ‘names can classify a person according to his or her sex, ethnic origin, family status, social status, and, in those societies which give different names to children at different stages of their life cycle, by age.’ The quotation comes from a book about names and naming in early modern England;11 the assumption is that different languages and cultures will fix on different classificatory features. Clearly from a Greek point of view classification according to sex is a primary consideration. Hence the creation of Athanippe, Theodora, Diodote.

10 The evidence is discussed in full in an early article of mine (‘Thessalian Patronymic Adjectives’, Glotta 46 (1968), 85–106), which obviously would need updating, though I still accept its conclusions. 11 S. Smith-Bannister, Names and Naming Patterns in England 1538–1700 (Oxford, 1997), 15.

The gender of an adjective (including a compound in adjectival function) is disambiguated by the syntax of the sentence in which it occurs, but this is not necessarily true for a personal name, certainly not true in one of the main uses of names, as forms of address,12 and not true in an important form of written document, the list. In the history of Greek over a long period, a com- bination of factors, including iconic factors, tends to introduce a formal marking of gender in all nominal inflection, but the phenomenon occurs first in personal names, so that the sex of individuals may be formally marked. A name like Athanippe is particularly striking if we remember that it is a compound of 4πποy and that 4πποy, which can be preceded by both the masculine and the feminine article, indicates both a stallion and a mare.13 The classificatory function of names also explains the peculiar ‘meaningful’ and ‘meaningless’ nature of some onomastic compounds. The primary function is not connotative or descriptive; in the so-called ‘irrational’ names the original function of composition, that which leads to the creation of the name, is not to describe the mental or physical characteristics of an individual but to identify him or her; in addition, the name may mark his or her position within a community or a kinship group.14 At a later stage the re-use of the ‘irrational’ name is determined by the desire to continue a name which was associated with a particular individual or, once again, a particular group. This lack of concern for connotative/descriptive meaning is also a precondition of the creation of abbreviated names, though it obviously does not prompt it.

12 Forms of address need not have been the primary motivation for the formal marking of gender; here too the obvious reference is to Socrates’ complaint in Aristophanes’ Clouds (685 ff.) that the vocative &Αµυνα sounds like a feminine. The point is that in forms of address the ‘addresser’ is in general likely to know the sex of the ‘addressee’. 13 Some women’s names in Greek are neuter, but this need not count as an exception to the general need for marking sex differences. Men’s names are never neuter. 14 I deliberately avoid the current philosophical discussion about the nature of proper names, but the contrast between connotation and denotation, hinted at in my text, refers to John Stuart Mill’s influential (even if endlessly criticized) distinction between denotation and connotation (see e.g. A System of Logic Rationative and Inductive (1843), in John Stuart Mill, Collected Works (Toronto and London, 1973), VII, I ii § 5). For Mill a connotative ‘name’ like ‘man’ denotes an indefinite number of individuals who possess certain attributes (corporeity, animal life, rationality, etc.); the name denotes the individuals but connotes the attributes. Proper names, by contrast, ‘are not connotative: they denote the individuals who are called by them; but they do not indicate or imply any attributes as belonging to those individuals’ (ib., 33). It follows for Mill that ‘whenever the names given to objects convey any information, that is, whenever they have properly any meaning, the meaning resides not in what they denote, but in what they connote. The only names of objects which connote nothing are proper names; and these have, strictly speaking, no signification.’ (ib., 34).

Use, abbreviated forms and archaic forms Names are widely used as forms of address and practical reasons may lead to the abbreviation of long compounds. Child-language is also relevant. The so-called Lallnamen (Mimi, Nini, Loulou, to take French examples), with their characteristic reduplication, originate in child-language and then remain in use for adults. Child-language may also be responsible for the frequent gemination which we find, for instance, in Italian names like Memmo (for Domenico), etc. The need for abbreviated forms is easy to understand, but the modality of the creation of Kurznamen is less clear. Greek has definite rules about the parts of the compound which can be lost, about the suffixes which can replace the second element of the compound, and so on: ∆εξ αy, ∆εξε!y, ∆εξαy, ∆ ξιy, ∆εξων are all found in the Peloponnese (LGPN IIIA s.vv.), and can plausibly be interpreted as abbreviated forms of names like ∆εξθεοy, ∆εξµαχοy, ∆ ξιπποy, ∆εξιφα´ νηy, etc. Possibly ∆ ξων and even ∆ ξιοy could be added to the list of abbreviated forms.15 Yet not everything is allowed: *Ξων,*&Εξαy would be unlikely.16 Admittedly we do not know for certain what is possible and what impossible in the formation of abbreviated names; further study is definitely needed. We know even less about how and when these rules were defined; clearly we must assume some complex processes of grammaticalization, but their origins are so far unknown. Three facts are important:

1 the rules for the creation of abbreviated forms are not the same in all dialects and at all periods. As we shall see, a form like ∆εξε!y is found once at Argos, while equivalent hypocoristics like ∆εξαy are more frequent. In contrast with the Peloponnese, Attic has only ∆ ξιy (as well as ∆ ξιοy and ∆ ξων), while the full compound names are the same as in the Peloponnese; 2 rules exist from our first evidence and we can establish such rules for Mycenaean;

15 A second-century Athenian called ∆ ξων is the father of a ∆εξοχοy (cf. LGPN II s. vv.), which confirms that synchronically a link existed. On the other hand, ∆ ξιοy could be a sobri- quet based on the adjective δεξιy. 16 The normal form of abbreviation preserves the first part of a compound (e.g. in ∆ ξιy mentioned above) and may or may not preserve the beginning of the second part (as in Πα´ τροκλοy vs. the original Πατροκλ 1ηy). There are rare examples of abbreviated forms which cut the first part of the compound; Masson, ZPE 37 (1980), 111 ff. OGS, 353 ff., supports the explanation of the rare masculine Χα´ ριy as an hypocoristic of compounds like &Ανδρχαριy, Ατχαριy, Τιµχαριy, etc.

3 abbreviated forms and non-abbreviated forms normally co-exist, but it is possible for the latter to replace the former in their entirety.

When a name exists both in its full compounded form and in one or more abbreviated forms, the choice of the one or the other form by the parents at birth must have been deliberate, even if later it is possible for the child to be known by more than one name. In general, a form of the name which soun- ded more archaic than the current one must have been due to a precise choice. In this way we account for the multiple continuations of a name like ÆDelpryd. Audrey seems to be the normal phonological development, but in a literate society (or for that matter in a society which is rich in oral poetic performances) older forms of words may be recorded in writing or in the poetic tradition. In the case of names, the incentive to resurrect them or to continue them in the original form may be stronger than for other lexical items. The choice of a specific form establishes a link with a specific person or period; in other words, once again it introduces a definite form of classification. Etheldreda is a typical product of nineteenth-century fashion.

The intentionality of naming and its consequences

A general point emerges from these apparently disconnected observations. I shall call it rather grandly the intentionality of naming. Speakers are more aware of what happens to names than of what happens to other lexical items. There are various reasons for this, including the fact that there are times in life when parents must choose names for their children. In the modern world publishers find it useful to keep in print dictionaries of personal names, which are avidly perused by expectant parents. This has consequences. If in the course of the history of English so¯na becomes soon just as mo¯na becomes moon and fo¯da becomes food, it seems likely that we are dealing with a phonetic change which is unconscious and mostly gradual. This type of unconscious phonological development also affects personal names, but for personal names speakers are capable of retrieving or preserving older forms. Consider, for instance, the name of King Nikoklewes of Paphos in Cyprus, who died c. 309 BC. He belongs to a period when intervocalic [w] tends to dis- appear. One of his syllabic inscriptions (ICS, no. 90) has the form ne-a-se (i.e. ν αy) and not the expected *ne-wa-se; in another (ICS, no. 91), his father is

given the title (in the genitive) of pa-si-le-o-se (i.e. βασιλ$οy) and not pa-si-le-wo-se. Yet all his syllabic inscriptions have the form ni-ko-ke-le-we-se or, in the genitive, ni-ko-ke-le-we-o-se (KOUKLIA, no. 237).17 To the linguist it looks as if phonological change has been deliberately suspended, but that is not the case; sound change goes on, but an older form of the name has been preserved or revived.18 The same principle may apply in the case of dialect shift. If a community, for instance, abandons its dialect in favour of a new standard (as the Greek regions did for the koine), two possibilities are available for proper names: they may be retained in the dialect form or they may be replaced by new names, which in their turn may be entirely new or may simply be the old names adapted to the new standard.19 Such decisions are not unconscious but deliberate. There are two consequences of this state of affairs: one more specific, the other of greater generality. First, it is now linguistically justified to use in the study of personal names data which stretch over a much longer period than is sometimes possible for other lexical items. Thus, for instance, scholars interested in the Greek dialects will stop studying Arcadian inscriptions after the third century or so since the dialect is no longer written, but in doing so they may miss the dialect information provided by the onomastics of the later inscriptions. A name like Καλλικρ τεια, with the archaic vocalism of the second element of the compound (-κρ τεια for -κρα´ τεια of the koine), is attested in the second/first century BC in Arcadia (LGPN IIIA s.v.) but shows an old pattern typical of the dialect. Secondly, a study of personal names may be highly revealing if we are interested in the cohesion and cultural continuity of a specific community. It tells us something not only about the natural preservation or otherwise of onomastic characters, but also about a set of deliberate choices in

17 The references are to O. Masson, Les inscriptions chypriotes syllabiques, 2nd edn (Paris, 1983) ICS; O.Masson and T.B. Mitford, Les inscriptions syllabiques de Kouklia-Paphos (Konstanz, 1986) KOUKLIA. 18 It is of course a moot point whether we are dealing with a fact of the written language or of the spoken language, and a decision may not be possible. 19 I owe to Dorothy Thompson the reference to a splendid article by Willy Clarysse (‘Ethnic Diversity and Dialect among the Greeks of Hellenistic Egypt’, in A.M.F.W. Verhoogt and S. P. Vleeming (eds), The Two Faces of Greco-Roman Egypt. Greek and Demotic and Greek-Demotic Texts and Studies Presented to P. W. Pestman (Leiden, 1998), 1–13), which traces the survival of Greek dialects in Egypt. Where I partially disagree with the method is when the author implies that the survival of names with non-Ionic–Attic long a¯ implies the survival of the dialect. This may well be the case but it would be conceivable that koine-speakers kept the old forms of the personal names.

name-giving and name-preserving that, in their turn, reflect specific attitudes to language but also to community life, kinship, continuity, etc. In a cultural context characterized by linguistic variety, such as that of pre-koine Greece, personal names may throw new light on the cultural diversity between different regions in a way which goes beyond the purely linguistic and merges with the analysis of attitudes or mentality.

Personal names and the koine

If these assumptions are correct, can we use them to enlighten real data? I now turn to more specific work which has recently become possible thanks to LGPN and to the accumulation of new data which it has provided. As in an earlier article,20 I concentrate on Arcadia but keep in mind data from other regions; I shall occasionally look at Cyprus because I subscribe to the fairly general view that the dialects of Cyprus and Arcadia continue a single original dialect. When the two dialects diverge this is doubly interesting because it is clearly due to an innovation. It goes without saying that most of what I have to say depends on the availability of the relevant volumes of the LGPN. It is well known that at some stage the dialectal variety of Greek is replaced —in writing at least—by a sort of standard language, the Attic-based koine. The exact period varies from region to region but at some point from the late fourth century onwards we begin to observe that inscriptions in dialect are replaced by inscriptions in koine.21 Often inscriptions in dialect and inscriptions in koine co-exist; sometimes the choice of language depends on the content or style of the inscription: public documents may be in koine while private documents are in the local dialect, but vice versa is also possible. How do we distinguish the koine from the local dialect? There are obviously a

20 A. Morpurgo Davies, ‘The Morphology of Personal Names in Mycenaean and Greek: Some Observations’, in Floreant Studia Mycenaea, Proceedings of the Tenth Mycenaean Colloquium (Vienna, 1999), 389–405. 21 It is not easy to define features and distribution of the koine. For an assessment of the prob- lems one may refer to the two volumes edited by Claude Brixhe, La Koiné grecque antique:I,une langue introuvable? (Nancy, 1993); La Koiné grecque antique: II, La Concurrence (Nancy, 1996); and to one of the concluding remarks of the article by Brixhe and Hodot in the first of these volumes (p. 20): ‘En définitive, la seule langue qui mérite réellement le nom de koiné est le reg- istre supérieur de la langue écrite (prose littéraire, documents diplomatiques, décrets munici- paux...).’

number of features which can be used, but the omnipresent one is phonological: the presence of a long e-vowel for an inherited long a (the contrast is that between Attic and Ionic µ+τηρ vs. µα´ τηρ of the other dialects). Onomastics may take part in the change, and suddenly in a given place we are presented with ∆ηµ+τριοy and Φιλδηµοy instead of the earlier (or contemporary) ∆αµα´ τριοy and Φιλδαµοy. The phenomenon naturally does not apply to Arcadia only and is generalizable for all non-Attic–Ionic areas. Yet there is regional diversity not only in the speed at which the koine progresses but also in the way in which it progresses: onomastics is a case in point.

The replacement of = with η in Arcadian names A glance at a list of Arcadian names reveals that forms with η for α are less frequent than one might have expected. I support this statement with the analysis of a few nominal types and of the figures for their attestation extracted from LGPN IIIA.

=-feminines The majority of feminine names end in -= in all dialects except for Ionic–Attic and the koine. If at any stage in Arcadia koine onomastics had taken over, we would expect a reasonable number of η-nominatives. In fact LGPN IIIA offers ninety-nine different Arcadian feminine names which end in -= vs. seventeen which end in -η. In themselves these figures are not significant because the α-names include names which in Attic and in the koine would also end in -α, either because the final vowel is short or because it is long but follows ε, ι, ρ (as in Μεγα´ κλεια, ∆ιονυσα, Κλεοπα´ τρα). We are then left with some forty-five names which could have been replaced by η- names but are not. Yet even this figure is not very significant because, naturally enough, many of the names belong to texts which are not written in koine. More striking is, on the one hand, the analysis of the seventeen η- names and, on the other, the distribution of some specific α-names. Of the η-names two are doubtful readings (9Ηδ!µ[η], Λυσι<ο>δκη); two are only attested in literature (&Αν!τη, ∆ηµοδκη listed under ∆αµοδκα in LGPN IIIA); two occur in elegiac distichs with clearly poetic language (Κλεαρστη, &Ιλη); and seven are very late and mostly belong to the second century AD

(Ζωσµη, Καλλιτ!χη, Νικοµα´ χη, Σγη, Φιλππη, Φοβη, Τ!χη; 9Ηδ!µ[η], if correct, and &Ιλη could also be mentioned here).; The remaining four names (Εα´ νθη, A Υλη, &Επιγνη and &Αθανππη) may require some additional discussion. Epigone is interesting: she is honoured as a benefactor by Mantinea–Antigoneia between the end of the first century BC and the beginning of the first century AD; the text (IG V (2) 268) starts in koina, i.e. it has =-forms like C πλιy, Dν ατE, but then includes a florid decree in Attic (or koine), which Wilamowitz thought was an example of Asianic style. All through, including the first part, the name is quoted as &Επιγνη. To judge from the three published volumes of LGPN, the name is late (imperial period) and does not normally appear with final -α.22 The obvious conclusion is that the woman in question was in fact called &Επιγνη, either because in Arcadia there was no equivalent &Επιγνα or because, if such a name existed, her parents and/or she herself preferred the η version. In either case the presence of this name with its η, both in the part of the text which refers to C πλιy in the koina and in that which refers to τFν πλιν in the koine, indicates that an alternation of dialect forms in a text does not automatically imply an alternation in the form of the names. AΥλη is even more problematic. The Greek evidence quoted for it in LGPN I–IIIA is limited to an occurrence in a text, dated to the third–second centuries BC, with only three words: &Επδοξον A Υλη χα0ρε; the first word is written somewhat differently and may have been added later. It is not certain, in my view, that we are dealing with a feminine nominative rather than with a masculine or feminine vocative, but if the name is a feminine, it may have been a literary name which conceivably was derived from a place-name.23

22 Pape-Benseler refers to three instances of &Επιγνα in Boeckh, CIG; in fact one (II 1957) is a wrong quotation; two (III 4287 and IV (sic but III) 6433) have the name in the η- form. There is, however, an instance of a first-century AD &Επιγνα in IG II2 9558 (CIG I 706) which most incongruously is the name of a Milesian woman, i.e. presumably of an Ionic woman for whom we would expect an η-name. If this is correct it demonstrates that an α-name existed, but offers no evidence for the fact that the name was created with an α- rather than an η-ending. It seems likely that the Arcadian woman in question was called &Επιγνη with a name which had reached Arcadia in that form. As for the other inscription the question remains of why a Milesian woman would have an α-name. 23 The text is edited by Te Riele, BCH 89 (1965), 584–9, who proposes two alternatives for &Επδοξον: either an honorific title like Gνδοξον or a woman’s name. However, if the word was added later, an interpretation ought to start with AΥλη χα0ρε; given the rarity of AΥλη as a personal name (as a city’s name the word is well known), various possibilities are open. It could be the vocative of a masculine name, in which case we could think of a non-attested *AΥληy,a possible Arcadian equivalent of *9Υλε!y, which is both an ethnic and a personal name

Practically nothing can be said about &Αθανππη of Stymphalos.24 It is one of only three -ιππη names attested once each in the Peloponnese; the other names are all late, and it is difficult to believe in an early date for &Αθανππη. The editor does not provide any information about date or letter-shape but simply says that we are dealing with a funerary stele; the second α of &Αθαν- in my view is not sufficient to define the inscription as hellenistic, and since we cannot check the reading it is difficult to trust the text. Finally, Εα´ νθη occurs once at Thelphousa in the second or first century BC25 and contrasts with Εα´ νθα found once at Megalopolis (IG V (2) 479, 9) presumably after the late third century BC. The name may well be literary in origin, and the -ανθη feminines are reasonably well attested in Attica from an early date.26 In general it is unlikely that we shall reach firm conclusions which provide a rationale for the use of these four names, if they are all real, but it seems clear that, before the late imperial period, there is no tendency in Arcadia to replace final -α with -η in the feminine names; if this happens at all it is exceptional.27

The δαµο- compounds The compounds of δα˜ µοy/δ$µοy are omnipresent in Greece; Ionic–Attic and the koine naturally use δηµο- forms. Arcadia has some thirty-nine different

(cf. Apollodorus 1. 8. 2 and 6) as well as a dog-name (cf. Xenophon, Cyn., 7. 5, and see Perpillou (below n. 45), 211); for the alternation see Arcado-Cypriot Iερ+y vs. Iερε!y. More plausibly perhaps we could think of a vocative of the koine equivalent of the masculine *AΥλαy, which is not frequent but is attested (Bechtel, HP, 577). A feminine AΥλη is very rare (Pape-Benseler s.v.; LGPN IIIA s.v. ) and I have found no evidence for a AΥλα form. The conclusion is that the interpretation of Te Riele’s text is not clear, nor is it clear why the name appears with a final -η;in any case we are not in a position to say that, in this instance, an Arcadian a¯-name has acquired a koine -η.

24 Quoted by A.K. Orlandos in Praktika 1929, 92; no epigraphical information is provided about the size of the monument, the lettering, etc. 25 SEG 11, 1125. 26 Cf. O. Masson, Rev. Phil. 58 (1984), 101 OGS, 453, with the reference to F. Bechtel, Die Attischen Frauennamen (Göttingen, 1902), 5. The formation is certainly attested but the feminine counterpart of the -α´ νθηy names ought to be -α´ νθεια, which is in fact well attested, or -ανθy and not -α´ νθα or -α´ νθη (the -α´ νθοy compounds seem to be late). We may wonder whether -α´ νθη is a secondary formation directly built on the -α´ νθηy names, though this would not explain the two examples of -α´ νθα names in LGPN IIIA. 27 Though I do not have as yet any clinching evidence, I would not exclude the possibility of the reverse phenomenon: an Attic name in which -η is replaced by -α to make the name more archaic. I do in fact wonder whether Εα´ νθα is a case in point.

compound names which have ∆αµ(ο)- as first element,28 and some thirty-one whose second element is -δαµοy. It has no instance of ∆ηµ(ο)- or -δηµοy compounds, though such compounds are in fact attested in other parts of the Peloponnese, and though names like ∆αµοκρα´ τηy, ∆αµοξ να, ∆αµστρατοy belong to our era.29

Derived names in -δαy Finally, a broad-range test can be conducted on the numerous names which end in -( )δηy in Attic–Ionic and in the koine, but in -δαy elsewhere (cf. Νικδαy, etc.). LGPN IIIA lists more than 130 Arcadian names of this type vs. only nine ending in -( )δηy. Of these, two occur in literary sources (&Ακεστορδηy,30 Πυλα´ δηy), one in an Egyptian papyrus (9Ηρακλεδηy), and one in an Egyptian inscription where it appears in the genitive only (&Ασκληπια´ δου), so that we cannot be absolutely certain about the nominative. The remaining six (&Αριστεδηy, ∆ιοσκουρδηy, 9Ηρωδηy, Λεωνδηy, Μιλτια´ δηy, Σωτα´ δηy (twice)) all belong to the second or third centuries AD. The exception is the third instance of the name Σωτα´ δηy which is reported to appear in an Attic inscription of the first century BC or first century AD where it refers to the father of a Λυκσκα, a woman from Alipheira. However, the name is in the genitive (Σωτα´ δου) and we cannot be certain of the nominative form.

Summary: the replacement of α with η in names A quick check of the other names (not many) which seem to have η where the dialect form would require α reveals a similar picture to that described above: most of them are only attested in literary sources, or outside Arcadia, or in very late inscriptions.31 The one interesting form is ΑJσχ!τηy, the name of a δαµιοργy from Kleitor listed in a proxeny from the fourth century (IG V (2) 1). It is the only name in a list of forty-nine which has an unexpected η. Hiller von Gaertringen accents ΑJσχυτ$y without an explanation; unless

28 I have excluded hypocoristics like ∆αµ αy, ∆α˜ µιy because I cannot entirely exclude that they belong to different roots; ∆αµα´ τριοy obviously does not belong here. 29 For ∆ηµοδκη, attested only in literature, see above under ‘a¯-feminines’. 30 Listed under &Ακεστορδαy in LGPN IIIA s.v. 31 Some of the apparent exceptions have already been mentioned. In addition I can only list from LGPN IIIA Βηθοy vs. Βοα´ θοοy (once in the imperial period, once in IG V (2) 190, a restored text from the second–first century BC with only this name); ∆ηµ+τριοy, once in the imperial period and once in a third–second-century text (SEG 39, 391), Νικηφροy, once in the imperial period and once in the second–first century BC in a single-word text (IG V (2) 197).

it is a misprint he may have thought of a contraction of - αy (otherwise not found in Arcadian). However, the name is found in Euboea and in Attica (I know of no evidence for ΑJσχ!ταy) and it is conceivable that the future demiourgos was given an Attic name. It is not conceivable that the names were automatically adapted to the koine since none of the others is. To put it more succinctly, Arcadia did not, before the second and third centuries AD, i.e. a few centuries after the introduction of the koine, modify its names to fit with the prevailing linguistic forms. A good demonstration of the phenomenon comes from IG V (2) 11, a proxeny decree for a Thessalian written in spotless koine in the second part of the third century BC. The first line has θεLy τ!χη and the last IερεMy τ$y &Αθηνα˜ y. Yet the list of στρατα- γο (sic) at the end has standard Arcadian names in the standard Arcadian forms; Εδοξδαy is not replaced by Εδοξδηy. In other words, the names are not adapted to the koine, nor is the title of the magistrates, but everything else is. Even more striking is the earlier proxeny IG V (2) 1 to which we have already referred. The proxeny is for an Athenian; it dates from 369–361 and is written once again in good Attic, but the last line of the decree before the list of magistrates is: δαµιοργοN ο4δε Oσαν. As in the later text the magistrates’ title is in the local form even if the verb is not. The names, as we have seen, are not in the Attic form (with the possible exception mentioned) and one of them still preserves a digamma. In other words, personal names and titles of magistrates in the fourth and third centuries are preserved in the original form even when the dialect gives way to Attic or the koine; in some sense, then, in Arcadia personal names and titles have the same ‘technical’ status.

Syllabic script vs. alphabetic writing in Cyprus We may well think that this tendency to preserve the original forms of names is normal, but the region which is linguistically close to Arcadia, Cyprus, does not support this view. There we have texts both in the local syllabic script and in the Greek alphabet. The former are regularly in the local dialect, the latter almost exclusively in the koine. Thus, in an honorific inscription dated by the editor to c. 190 BC,32 where it would be conceivable to have traces at least of the epichoric dialect, we have instead:

32 T.B. Mitford, The Inscriptions of Kourion (Philadelphia, 1971), no. 49.

Φιλ0νοy καN Μ ντωρ καN &Ον+σιλοy Μ ντορα τLν QαυτRν πατ ρα &Απλλωνι ν ν 9Υλα´ τηι

By contrast the epichoric forms of both the name (Ona¯silos) and the god’s epitheton (Hylata¯s) are preserved in the syllabic inscriptions.33 The use of koine forms is not limited to second-century texts; in the alphabetic part of one of the rare digraphic and bilingual syllabic inscriptions from fourth-century Amathous we find: 9Η πλιy S &Αµαθουσων &Αρστωνα &Αριστνακτοy επατρδην.34 But the most striking texts are the digraphic dedications from Kafizin. These are on pots offered by artisans to the Nymphaeum of Kafizin in east Cyprus at the end of the third century BC.35 We have sixty-six syllabic texts and two hundred alphabetic texts including some thirty-two digraphic texts. Once again the alphabetic texts are all in koine, the syllabic texts in the local dialect with clear koine influence. However, as far as the clearest marker of the koine is concerned, viz. the presence of η where the local dialect would have [a¯], the two forms of writing are internally consistent: the syllabic texts have only -a- and the alphabetic texts have only η or the ε with which occasionally it alternates. This applies to all words concerned, including gods’ names and personal names. The dedications are made to nu-pa-i or nu-mu-pa-i in the syllabic script but to Ν!µφη(ι) (once Ν!µφει) in the alphabetic. The recurrent name of the δεκατηφροy / δεκατεφροy or te-ka-ta-po-ro-se is &Ονησαγρα in the alphabetic but o-na-sa-ko-ra-se or o-na-a-ko-ra-se in the syllabic script. After Kafizin the syllabic script disappears and the alphabetic forms with η take over. The contrast with Arcadia could not be greater.36

33 Cf. A. Hintze, A Lexicon to the Cyprian Syllabic Inscriptions (Hamburg, 1993), s.vv. o-na-si- lo-se, u-la-ta-i; M. Egetmeyer, Wörterbuch zu den Inschriften im kyprischen Syllabar (Berlin, 1992), s. vv. o-na-si-lo, etc.; u-la-ta-i, etc. 34 Masson, ICS, no. 196. The syllabic text is in Eteocypriot, so that we have a difference both of language and of script. 35 For the texts see T.B.Mitford, The Nymphaeum of Kafizin. The Inscribed Pottery, Suppl. II to Kadmos (Berlin and New York, 1980); O. Masson, ‘A propos des inscriptions chypriotes de Kafizin’, BCH 105 (1981), 623–49; C. Consani, Persistenza dialettale e diffusione della KOINH a Cipro (Pisa, 1986). 36 One should note, however, that not all alphabetic names have η where the koine would have it. Below (n. 40) I refer to an alphabetic Στασι- vs. the expected (and more frequent) Στησι-; similarly, while a stem &Ονησι- is frequent in the alphabetic texts, we also find (more rarely) alphabetic names which start with &Ονασι-. The general rule, nevertheless, is that the syllabic script has no names with -e¯- from -a¯-, while the alphabetic inscriptions mostly (though not always) have -η- from -a¯.

Arcadia vs. Cyprus The evidence from Cyprus is not so easy to interpret, since the role of the two scripts is still imperfectly understood and the contrast is unparalleled elsewhere in Greece. Yet all the evidence which we have, and the Kafizin evidence in particular, tells us that dialect-speakers at some (late) stage had acquired a set of mental equivalences between dialect and koine forms, and in the written language at least they were prompted by the script used to choose one or the other member of the equation. We cannot know what happened in the spoken language. In Arcadia, however, the position is different: apparently an Arcadian writer and presumably a speaker had to decide whether the dialect or the koine were the preferred means of expression, but for personal names they had little choice: they preserved the original form. And yet there must have been a number of names, particularly compounds, which were wholly transparent. Anyone proficient in the koine would have known suit- able equivalents: an Arcadian name like Φιλδαµοy is a transparent compound of δα˜ µοy. There are plenty of Arcadian texts in which δ$µοy, the koine form, is preferred, but even so Φιλδαµοy keeps its -α- form. The data then support the view that in Arcadia personal names had a special status due to their function, and that this special status is responsible for the linguistic conservatism which we observe; at the same time it becomes clear, partly because of the contrast with Cyprus, that this special status is a cultural phenomenon which cannot be generalized from region to region.

The -κρ τηy/-κρα´ τηy compounds in Arcadia and Cyprus Those who find it difficult to understand the consistency of some Cypriot spellings with η (ε) in the alphabetic texts vs. -α- of the syllabary may wish to argue that we are not dealing with conscious decisions, but simply with writing or spelling techniques. Just as a French-speaker knows that in the written language the plural of homme (and numerous other words) is written with a final -s (hommes) even if this is not pronounced, or just as non-native speakers of English are taught to write the singular man differently from the plural men even if often they cannot hear that they are pronounced differently, so a Cypriot may have been taught to write da-mo in the syllabic script and δ$µοy in the alphabet, without necessarily pronouncing the two vowels differently. The suggestion is counter-intuitive (largely because of the alter- nations η/ε in the alphabetic forms), but in any case all doubts are dispelled

by the analysis of other instances of differing alphabetic and syllabic texts for which a similar explanation based on spelling is not possible. In the early Arcadian and Cypriot inscriptions (as well as in the Lesbian inscriptions) the standard Greek compound names of the Socrates type appear with a second element -κρ τηy instead of -κρα´ τηy. This is the older form of these compounds; cf., for example, the numerous Homeric compounds in -πενθ+y (πολυπενθ+y, etc. ) next to the unique αJνοπαθ+y, which exemplifies the pattern which will dominate later.37 In both Arcadian and Cypriot we tend to find -κρ τηy in the earlier texts, while the late texts tend to have -κρα´ τηy. Yet we are not necessarily dealing here with an intrusion of the koine. The process through which -κρ τηy is replaced by -κρα´ τηy, and the noun κρ τοy (attested in Alcaeus) is replaced by κρα´ τοy, may well have happened independently in the various dialects; the model is provided by adjectives like κρατ!y, κρατερy where the -α- vocalism is expected. The distribution of -κρ τηy/-κρα´ τηy in personal names is nevertheless interesting. LGPN I lists for Cyprus twenty-six different -κρα´ τηy names, including Κρα´ τηy (which is presumably a back-formation from the compounds), and eleven different -κρ τηy names; new readings etc. bring the number of -κρ τηy names to fourteen or even fifteen.38 In Arcadia there are twenty-nine different -κρα´ τηy names vs. fifteen different -κρ τηy names. The former are distributed through all periods, the latter start in the fifth century but do not go beyond the hellenistic period. Dr Torsten Meißner has pointed out to me that in IG V (2) 38, a long third-century Tegean catalogue of citizens and metics, the two metics with this type of name have -κρα´ τηy names (Σωσικρα´ τηy and ∆εξικρα´ τηy) while two of the eponymous priests and one citizen have -κρ τηy names (Σωκρ τηy, Τιµοκρ τηy, Ερυκρ τηy), and no citizen or priest has a -κρα´ τηy name. In other words, the -κρ τηy

37 Cf. e.g. E. Risch, Wortbildung der homerischen Sprache (Berlin, 1974), 78 f., 82. 38 The Cypriot names in -krete¯s are: Aristokretes (syll.) vs. &Αριστοκρα´ τηy; Menokrete¯s vs. Menokrate¯s (both syll.); Minokrete¯s (syll.); Ona¯sikrete¯s (syll.) vs. &Ονησικρ τηy (o-na- si-ka[ in ICS, no. 246, 3 could be Ona¯sikrate¯s but also Ona¯sikharis); Pasikrete¯s (syll.) vs. Πασικρα´ τηy ; Philokrete¯s (syll.) vs. Φιλοκρα´ τηy;So¯krete¯s (syll.) vs. Σωκρα´ τηy; Sta¯sikrete¯s (syll.) vs. Sta¯sikrate¯s (x2, syll.), Στασικρα´ τηy, Στησικρ τηy; Timokrete¯s (syll.) vs. Τιµοκρ τηy; Timukrete¯s (syll.); Zo¯wokre[te¯s (syll.). The names in question are fewer if we identify Menokrete¯s with Minokrete¯s and Timokrete¯s with Timukrete¯s (all of them written in the syllabic script). Additional names to those listed in LGPN I are the somewhat fragmentary Kuprokrete¯s and Wikrete¯s from syllabic texts of Golgoi (Masson, Kadmos 18 (1989), 159, 162), the compound ]tokrete¯s (perhaps Pnutokrete¯s) from Abydos (Masson, ICS, no. 409), and the fragmentary genitive k]reteos, also syllabic, from early Kouklia (KOUKLIA, no. 1).

names are likely to be the traditional ones and there may be, though so far this has not been demonstrated, a social class distinction between -κρ τηy and -κρα´ τηy names. The latter eventually prevail. In Cyprus, however, there is a further distinction: the syllabic texts have a clear preference for -krete¯s but also have some examples of -krate¯s (Menokrate¯s and Sta¯sikrate¯s), while the alphabetic texts have no examples of -krete¯s at all.39 Here we cannot explain the presence of -κρα´ τηy in the alphabetic texts as due to purely graphic factors (the Cypriots pronounced [krete¯s] but learned to write -κρα´ τηy in the alphabet), since the syllabic texts make it clear that all Cypriots must have known both, for instance Stasikrete¯s and Stasikrate¯s, and that the syllabary could indicate the difference. It looks as if, once again, the Cypriots simply adopted for the alphabetic texts the koine form of their names.40

Arcadian vs. Cypriot attitudes and their explanation In contrasting the Arcadian and the Cypriot attitude one may well ask whether the difference is partly the result of the different ethnic situations in Arcadia and in Cyprus: on the one hand, a fundamentally homogeneous linguistic area and on the other, an island where at least three different and non-mutually intelligible languages were spoken—Greek (both as Cypriot and as koine at a later stage), Phoenician and Eteocypriot. Onomastic adap- tation from one language to the other is documented for both Phoenician and Eteocypriot and is indeed what we would expect.41 The extension of the same principle to the local dialect and the koine should not be surprising.

39 There is, however, a feminine Κρετ (the daughter of a Κλ ων) at the end of the third century BC (LGPN I s.v.). 40 An interesting case is that of the syllabic name Stasikrete¯s, which is also written in the syllabary as Stasikrate¯s, while in the alphabet it has the forms Στασικρα´ τηy and Στησικρ τηy. The form with η in the first syllable is expected in the alphabetic writing while the preservation of = in Στ=σικρα´ τηy is unexpected, but not unparalleled, in the word-stem (as contrasted with the ending). 41 For Phoenician cf. e.g. ICS, nos 215–16 from fourth-century Tamassos, where the Phoenician names mnh.mbnbnh.dˇs and ‘bdssm are rendered with ma-na-se-se o-no-me-ni-o-ne and a-pa-so-so-mo-se, i.e. Mnase¯s ho No¯smenio¯n Apsaso¯mos respectively. Mnases’ father carries a Greek name which is obviously a calque from the Phoenician. Masson thought that Mnase¯s was a second name (Greek in origin) carried by mnh.m, but other interpretations would be possible. For Eteocypriot cf. e.g. ICS, no. 196 from Amathous (fourth century), where the Greek &Αρστων &Αριστνακτοy corresponds to a-ri-si-to-no-se a-ra-to-wa-na-ka-so-ko-o-se, and see Masson’s commentary.

Continuity and discontinuity in personal names

Both in the title of this paper and in parts of what precedes reference is made to linguistic continuity as shown by onomastics or, more specifically, by personal names. The fact that place-names survive changes in culture and even language is well documented; for personal names the position is somewhat different, but the evidence adduced ought to demonstrate that, owing to the way in which they are transmitted, personal names may show a higher level of formal continuity than other elements of the lexicon. In what follows I would like to return to some work which I have done in recent years about Arcadian and Mycenaean names and to consider, first, an example of discontinuity and second, an example of apparent discontinuity which in fact turns out to provide evidence for structural continuity.42

The ε!y-names We associate names like &Οδυσσε!y, &Αχιλλε!y etc. with epic language and with heroic personalities: there are some fifty such names in Homer. Yetnames of this type play a very limited, practically non-existent role in Greek onomastics. Thus, to give some figures, of the 8,306 different names listed in LGPN II (Attica) only sixty-nine (i.e. 0.83 %) end in -ε!y,inLGPN I 0.69 % (ninety-eight out of 14,150), and in LGPN IIIA 0.77 % (eighty-one out of 10,423).43 In Arcadia, to come down to lower figures, five names out of 1,880 (i.e. 0.26%) end in -ε!y. If, instead of looking at different names, we look at different individuals who have an ε!y-name, the picture is not very different; in Attica, out of 62,360 individuals whose name we know, some 236 (i.e. 0.38 %) have an ε!y-name. An obvious reaction is to treat the Homeric names as purely poetic invention and to assume that Greek parents did not like poetic names, but this is no longer possible after the decipherment of Linear B. In Mycenaean there are at least 130 different eus-names out of c. 1,800 names, i.e 7.26 %. These are names of real individuals, not heroes and not gods. It follows that from this point of view the onomastics of epic language matches to a certain extent that of Mycenaean, but the later language is different.44

42 Cf. A. Morpurgo Davies, ‘The Morphology of Personal Names in Mycenaean and Greek’ (above n. 20). 43 Note that Αλεξε!& y quoted in LGPN IIIA is a vox nullius and must be read ∆εξε!y; see below n. 50. 44 For part at least of its evidence. Perpillou, Les substantifs grec en -ε!y, e.g. at 223 f. is aware of the scarcity of the post-Mycenaean data, but without the volumes of LGPN, the contrast between Mycenaean and alphabetic documentation was not quantifiable.

There is discontinuity here and we would like to know what prompted it. It cannot be due to phonological or morphological considerations because, for instance, ethnics in -ε!y and common nouns in -ε!y survive and to a certain extent multiply.45 Heroic epic names may have been avoided, but it would have been possible to create new names to compensate for this gap.46 Why this did not happen – or did not happen frequently enough – remains mysterious.

Hypocoristics in -σε!y In spite of the gap mentioned above, there are some other forms of continuity which involve ε!y-names. One of the functions which -eus has in Mycenaean is that of creating hypocoristics from personal names built on compounds of the so-called τερψµβροτοy-type. A name like Mycenaean a-re-ke-se-u , i.e. Alexeus, is in all likelihood an abbreviated form of names like &Αλ ξανδροy, &Αλεξµαχοy, etc.47 Similar formations are de-ke-se-u : Dexeus, ka-ri-se-u : Khariseus, o-na-se-u : Onaseus, pa-ra-ke-se-we (dative) : Praxeus.48 The type in general has disappeared from Greek, but there are iso- lated survivals, as pointed out by Masson:49 Μνησε!y occurs in Plato and twice in inscriptions, once in Chios and once in Chalcis; ∆εξε!y occurs once, or possibly twice, at Argos.50

45 The standard work of reference is J.L. Perpillou, Les substantifs grecs en -ε!y (Paris, 1973). The productivity of the suffix increases in the classical and immediately post-classical periods but seems to diminish in our era. 46 In the late imperial period some of the heroic names become fashionable again: the twenty- seven Athenians called Θησε!y all belong to the second or third century AD, as do the ten called Νηρε!y and the five called Περσε!y. Similarly, in the three volumes of LGPN published so far the attestations for &Αχιλλε!y are mostly very late. 47 a-re-ka-sa-da-ra, i.e. &Αλεξα´ νδρα, is actually attested in Mycenaean. 48 Note that strictly speaking forms like &Αλ ξανδροy, &Αλ ξιπποy, &Αλεξµαχοy, etc., which lie behind (are at the origin of) hypocoristics like &Αλεξε!y or ,Αλεξιy, do not count as τερψµβροτοy-compounds since they are built on the stem of -λ ξω. A similar formation is probably that of Mycenaean a-we-ke-se-u, which is normally understood as Awekseus and compared with Homeric -(1) ξω and Attic αξω, αξα´ νω. Mycenaean has a large number of names ending in -se-u (listed by Perpillou, 222 f.) but it is far from clear that they are all abbreviated forms of compounds. For the names quoted above we may refer to compounds such as ∆ ξανδροy, ∆εξµαχοy, ∆εξικρα´ τηy, etc. or &Ονα´ σανδροy, &Ονασικρα´ τηy, &Ονασικλ$y, etc., or Πραξαγραy, Πραξδαµοy, Πραξιτ ληy, etc. 49 SMEA 2 (1967), 36 ff. OGS, 106 ff.; cf. also Perpillou, 223 f. 50 Masson, loc. cit., 37 OGS, 107 points out that the Argive fourth-century inscription (CEG 2, 816), where the editor read &Αλεξε!y, in fact has the name ∆εξε!y. &Αλεξε!y should therefore be removed from LGPN IIIA. LGPN IIIA also mentions an unpublished instance of ∆εξε!y, also from Argos.

The more normal hypocoristics from this type of compounds are i-stems like ,Αλεξιy, ∆ ξιy, ,Ονασιy, Ζε ξιy, etc., a type unknown to Mycenaean. Arcadian is rich in hypocoristics which end in -ιy (,Αρετιy father of ,Αρετακλ$y, Εθιy vs. e.g. Εθ!δαµοy, Εθυκλ$y, etc.) but has no -σιy forms nor, as we have seen, -σε!y forms. This does not exclude the presence of hypocoristics; there are no less than three suffixes which are exploited for this purpose: -ων,-αy and - αy. Thus, in correspondence with Μνασλαοy, Μνασδαµοy, Μνασστρατοy, Μνασιτ ληy Arcadian does not have Μνα˜ σιy, but has Μνασων, Μνασαy and Μνασ αy (cf. LGPN IIIA s.vv.). If we add up all the instances, the following list gives an idea of the frequency of different names formed in this way:

-σαy/-ξαy: twenty different hypocoristics (e.g. &Αγασαy, &Αλεξαy, ∆εξαy, Πραξαy, Μνασαy, Πασαy, Παυσαy); -σων/-ξων: ten different hypocoristics (e.g. &Αλεξων, Πραξων, Μνασων, Παυσων); -σ αy (no -ξ αy): six different hypocoristics ( &Αγασ αy, Γνωσ αy, Θρασ αy, Μνασ αy, Μοισ αy, Πασ αy).51

Once again Mycenaean Alexeus, Praxeus, Dexeus should be compared with &Αλεξαy, Πραξαy, ∆εξαy. There is no clear evidence that allows us to establish whether the three hypocoristic types arose at different times. It would be possible to suggest that -αy is derived from - αy, since in Arcadian [e] is often raised before a central vowel, but this too cannot be certain and the two suffixes may have a separate origin. One could also assume that the least frequent suffix is perhaps the one on its way out and consequently the oldest. If so, we may wish to argue that in Arcadian - αy replaced the -ε!y suffix which was used in Mycenaean to form hypocoristics from τερψµβροτοy-compounds. This is speculation. Further speculation might remind us of the existence of an -e- wa suffix in Mycenaean which could count as the antecedent of - αy and

51 Note that not all these names are strictly speaking hypocoristics of τερψµβροτοy-compounds, which have a verbal first element. I suspect, however, that synchronically all compounds with internal -σι-/-ξι- before the second element were treated in the same way from the point of view of word-formation. Hence all compounds which began with Πασι- or Θρασι- (like Πα´ σιπποy or Θρα´ σιπποy) behaved in the same way as e.g. Πα!σιπποy or Μνα´ σιπποy. One name may not belong here, Μοισ αy of Thelphousa, where the root points to a literary form of the word; see Masson, REG 94 (1981), 543 = OGS, 390.

might plausibly be taken as a derivative of -ε!y. One small bit of evidence may support a link between the two suffixes. Pausanias (6. 11. 4) refers to a ∆ροµε!y from Mantinea who was Olympic victor in 480 BC; about the same period one of the oldest Mantinean inscriptions, concerning a crime in the temple (IG V (2) 262), refers to a ∆ροµ αy. Could this be the same name or even the same person ?52 Whatever the answer, it looks as if Arcadia created hypocoristics from τερψµβροτοy-compounds or their equivalents in a way which avoids the usual -σιy termination and exploits instead - αy,-αy,-ων. The crucial point is that Arcadian, like Mycenaean, treats these compounds somewhat differently from the others and uses a special suffix to form their hypocoristics. Cypriot , on the other hand, has a name like ,Ονασιy which is clearly based on compounds of the &Ονασκυπροy type. From this point of view we find, once again, conservatism in Arcadia and innovation in Cyprus.

Conclusions

Linguistic conservatism and cultural or institutional or political conservatism do not always go hand in hand. Much depends on what attitude the members of a specific group take to their own language, and on how much they are aware of linguistic facts. Onomastics creates further difficulties; it is conceivable that speakers who take a cavalier, non-purist attitude to their language are nevertheless conservative in their onomastics—or, of course, vice versa. A rapid semantic shift in the lexicon is often a sign of rapid cultural change, but this may happen without any clear impact on personal names. On the other hand, a change in personal names also indicates sub- stantial cultural change, though not necessarily of the same nature. The decision to preserve or abandon one’s own family names or their traditional form is normally deliberate and is significant. That the Arcadians started to write in koine but did not choose to alter the title of their magistrates or the form of their personal names is important for an understanding of how the

52 The origin of - αy is far from clear, partly because we do not know if we can or cannot take it back to an [-e¯a¯s] form, with a long e-vowel which would then speak for a straight derivation from the eu-stems where the e-vowel was originally long (cf. Homeric βασιλ$οy, etc.). The fact that we know of a third-century Arcadian called Θηρ+αy (FD III (1) 83,14) with a name which may match exactly the Mycenaean name qe-re-wa, may or may not be significant.

Arcadians saw themselves; the different choice made by the Cypriots, who linked the form of their personal names, like that of other lexical items, to the script in which they were writing, is equally significant and requires interpretation. The linguist who investigates these matters cannot be given a simple set of instructions, but must observe and interpret each case in its own terms, remembering that a purely linguistic approach may not be sufficient. Thanks to the Lexicon of Greek Personal Names we now begin to have the basic data for a study which can both go deep and range widely; it is our good fortune that we can witness, and perhaps participate in, the opening of a new field of enquiry.

BE J. and L. Robert and others, Bulletin Épigraphique (in Revue des Études grecques, 1938– ) Bechtel, HP F. Bechtel, Die historischen Personennamen des Griechischen bis zur Kaiserzeit (Halle, 1917) CEG P.A. Hansen, Carmina Epigraphica Graeca, vol. 1, saeculorum VIII–V a. Chr. n.; vol. 2, saeculi IV a. Chr. n. (Berlin, 1983, 1989) FD Fouilles de Delphes 1– (Paris, 1909– ) Hatzopoulos, Macedonian Institutions M.B. Hatzopoulos, Macedonian Institutions under the Kings, 2 vols (Meletemata 22; Athens, 1996) Hatzopoulos-Loukopoulou, Recherches M.B. Hatzopoulos and L. Loukopoulou, Recherches sur les marches orientales des Téménides, i (Meletemata 11; Athens, 1992) Letronne, Oeuvres choisies Oeuvres choisies de J.-A. Letronne, assemblées, mises en ordre et augmentées d’un index par E. Fagnan (Paris, 1881–5: 1 sér. Égypte ancienne, 2 vols, 1881; 2 sér. Géographie et cosmographie, 2 vols, 1883; 3 sér. Archéologie et philologie, 2 vols, 1883–5) LGPN A Lexicon of Greek Personal Names I, The Aegean Islands, Cyprus and Cyrenaica, eds P.M. Fraser and E. Matthews (Oxford, 1987); II, Attica, eds M.G. Osborne and S.G. Byrne (Oxford, 1994); IIIA, The Peloponnese, Western Greece, Sicily and Magna Graecia, eds P.M. Fraser and E. Matthews (Oxford, 1997); IIIB, Central Greece, eds P.M. Fraser and E. Matthews (forthcoming, 2000) LIMC Lexicon iconographicum mythologiae classicae (Zurich and Munich, 1981–97) LSAG2 L.H. Jeffery, The Local Scripts of Archaic Greece, 2nd edn, with sup- plement by A.W. Johnston (Oxford, 1990) ML R. Meiggs and D. Lewis, A Selection of Greek Historical Inscriptions to the End of the Fifth Century BC, revised edn (Oxford, 1988) OCD3 S. Hornblower and A. Spawforth (eds), Oxford Classical Dictionary, 3rd edn (Oxford, 1996) OGS O. Masson, Onomastica Graeca Selecta, ed. C. Dobias and L. Dubois, 2 vols (Paris, 1990) Osborne, Naturalization 3–4 M. J. Osborne, Naturalization in Athens, 3-4 (Brussels, 1983)

PA J. Kirchner, Prosopographia Attica, 2 vols (Berlin, 1901) Pape-Benseler W. Pape and G.E. Benseler, Wörterbuch der griechischen Eigennamen (Braunschweig, 1863–70) Parker, Athenian Religion R. Parker, Athenian Religion: A History (Oxford, 1996) Robert, OMS L. Robert, Opera Minora Selecta: Épigraphie et antiquités grecques, 7 vols (Amsterdam, 1969–90) SEG Supplementum Epigraphicum Graecum, eds J.E.E. Hondius and A.G. Woodhead, 1–25 (Leiden, 1923–71); eds H.W. Pleket and R.S. Stroud, 26–7 (Alphen, 1979–80), 28– (Amsterdam, 1982– ) SGDI H. Collitz, F. Bechtel and others, Sammlung der griechischen Dialekt- Inschriften, 4 vols (Göttingen, 1884–1915) Sittig E. Sittig, De Graecorum nominibus theophoris (diss. Halle, 1911)